Performance analysis and linear optimization modeling of all-to-all collective communication algorithms

Hyacinthe Nzigou Mamadou, Guilherme De Melo Baptista Domingues, Takeshi Nanri, Kazuaki Murakami

研究成果: 著書/レポートタイプへの貢献会議での発言

2 引用 (Scopus)

抄録

The performance of collective communication operations still represents a critical issue for high performance computing systems. Users of parallel machines need to have a good grasp of how different communication patterns and styles affect the performance of message-passing applications. This paper reports our contribution of the analysis of collective communication algorithms in the context of MPI programming paradigm by extending a standard pointto-point communication model, which is P-LogP. We focus on MPI Alltoall since this function is one of the most communication intensive collective operations known. In order to reduce the gap between the predicted and the measured run-time, all the system parameters are also taken into account with the total performance estimation, by applying the linear regression modeling with the empirical data. Results on InfiniBand clusters show that the final performance prediction models can accurately capture the entire system communication behavior of all algorithms, even for large size messages and large number of processors.

元の言語英語
ホスト出版物のタイトルProceedings - 19th International Symposium on Computer Architecture and High Performance Computing, SBAC-PAD
ページ203-210
ページ数8
DOI
出版物ステータス出版済み - 12 1 2007
イベント19th International Symposium on Computer Architecture and High Performance Computing, SBAC-PAD - Gramado, RS, ブラジル
継続期間: 10 24 200710 27 2007

出版物シリーズ

名前Proceedings - Symposium on Computer Architecture and High Performance Computing
ISSN(印刷物)1550-6533

その他

その他19th International Symposium on Computer Architecture and High Performance Computing, SBAC-PAD
ブラジル
Gramado, RS
期間10/24/0710/27/07

Fingerprint

Communication
Message passing
Linear regression

All Science Journal Classification (ASJC) codes

  • Engineering(all)

これを引用

Mamadou, H. N., Domingues, G. D. M. B., Nanri, T., & Murakami, K. (2007). Performance analysis and linear optimization modeling of all-to-all collective communication algorithms. : Proceedings - 19th International Symposium on Computer Architecture and High Performance Computing, SBAC-PAD (pp. 203-210). [4384059] (Proceedings - Symposium on Computer Architecture and High Performance Computing). https://doi.org/10.1109/SBAC-PAD.2007.34

Performance analysis and linear optimization modeling of all-to-all collective communication algorithms. / Mamadou, Hyacinthe Nzigou; Domingues, Guilherme De Melo Baptista; Nanri, Takeshi; Murakami, Kazuaki.

Proceedings - 19th International Symposium on Computer Architecture and High Performance Computing, SBAC-PAD. 2007. p. 203-210 4384059 (Proceedings - Symposium on Computer Architecture and High Performance Computing).

研究成果: 著書/レポートタイプへの貢献会議での発言

Mamadou, HN, Domingues, GDMB, Nanri, T & Murakami, K 2007, Performance analysis and linear optimization modeling of all-to-all collective communication algorithms. : Proceedings - 19th International Symposium on Computer Architecture and High Performance Computing, SBAC-PAD., 4384059, Proceedings - Symposium on Computer Architecture and High Performance Computing, pp. 203-210, 19th International Symposium on Computer Architecture and High Performance Computing, SBAC-PAD, Gramado, RS, ブラジル, 10/24/07. https://doi.org/10.1109/SBAC-PAD.2007.34
Mamadou HN, Domingues GDMB, Nanri T, Murakami K. Performance analysis and linear optimization modeling of all-to-all collective communication algorithms. : Proceedings - 19th International Symposium on Computer Architecture and High Performance Computing, SBAC-PAD. 2007. p. 203-210. 4384059. (Proceedings - Symposium on Computer Architecture and High Performance Computing). https://doi.org/10.1109/SBAC-PAD.2007.34
Mamadou, Hyacinthe Nzigou ; Domingues, Guilherme De Melo Baptista ; Nanri, Takeshi ; Murakami, Kazuaki. / Performance analysis and linear optimization modeling of all-to-all collective communication algorithms. Proceedings - 19th International Symposium on Computer Architecture and High Performance Computing, SBAC-PAD. 2007. pp. 203-210 (Proceedings - Symposium on Computer Architecture and High Performance Computing).
@inproceedings{173a06681f9e4643aa9679afe9944069,
title = "Performance analysis and linear optimization modeling of all-to-all collective communication algorithms",
abstract = "The performance of collective communication operations still represents a critical issue for high performance computing systems. Users of parallel machines need to have a good grasp of how different communication patterns and styles affect the performance of message-passing applications. This paper reports our contribution of the analysis of collective communication algorithms in the context of MPI programming paradigm by extending a standard pointto-point communication model, which is P-LogP. We focus on MPI Alltoall since this function is one of the most communication intensive collective operations known. In order to reduce the gap between the predicted and the measured run-time, all the system parameters are also taken into account with the total performance estimation, by applying the linear regression modeling with the empirical data. Results on InfiniBand clusters show that the final performance prediction models can accurately capture the entire system communication behavior of all algorithms, even for large size messages and large number of processors.",
author = "Mamadou, {Hyacinthe Nzigou} and Domingues, {Guilherme De Melo Baptista} and Takeshi Nanri and Kazuaki Murakami",
year = "2007",
month = "12",
day = "1",
doi = "10.1109/SBAC-PAD.2007.34",
language = "English",
isbn = "9780769530147",
series = "Proceedings - Symposium on Computer Architecture and High Performance Computing",
pages = "203--210",
booktitle = "Proceedings - 19th International Symposium on Computer Architecture and High Performance Computing, SBAC-PAD",

}

TY - GEN

T1 - Performance analysis and linear optimization modeling of all-to-all collective communication algorithms

AU - Mamadou, Hyacinthe Nzigou

AU - Domingues, Guilherme De Melo Baptista

AU - Nanri, Takeshi

AU - Murakami, Kazuaki

PY - 2007/12/1

Y1 - 2007/12/1

N2 - The performance of collective communication operations still represents a critical issue for high performance computing systems. Users of parallel machines need to have a good grasp of how different communication patterns and styles affect the performance of message-passing applications. This paper reports our contribution of the analysis of collective communication algorithms in the context of MPI programming paradigm by extending a standard pointto-point communication model, which is P-LogP. We focus on MPI Alltoall since this function is one of the most communication intensive collective operations known. In order to reduce the gap between the predicted and the measured run-time, all the system parameters are also taken into account with the total performance estimation, by applying the linear regression modeling with the empirical data. Results on InfiniBand clusters show that the final performance prediction models can accurately capture the entire system communication behavior of all algorithms, even for large size messages and large number of processors.

AB - The performance of collective communication operations still represents a critical issue for high performance computing systems. Users of parallel machines need to have a good grasp of how different communication patterns and styles affect the performance of message-passing applications. This paper reports our contribution of the analysis of collective communication algorithms in the context of MPI programming paradigm by extending a standard pointto-point communication model, which is P-LogP. We focus on MPI Alltoall since this function is one of the most communication intensive collective operations known. In order to reduce the gap between the predicted and the measured run-time, all the system parameters are also taken into account with the total performance estimation, by applying the linear regression modeling with the empirical data. Results on InfiniBand clusters show that the final performance prediction models can accurately capture the entire system communication behavior of all algorithms, even for large size messages and large number of processors.

UR - http://www.scopus.com/inward/record.url?scp=47249095739&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=47249095739&partnerID=8YFLogxK

U2 - 10.1109/SBAC-PAD.2007.34

DO - 10.1109/SBAC-PAD.2007.34

M3 - Conference contribution

AN - SCOPUS:47249095739

SN - 9780769530147

T3 - Proceedings - Symposium on Computer Architecture and High Performance Computing

SP - 203

EP - 210

BT - Proceedings - 19th International Symposium on Computer Architecture and High Performance Computing, SBAC-PAD

ER -